Introduction
The retreat of sea ice in the Arctic this summer far exceeded expectations1, and the rate of meltback appears to be accelerating as the ice-free ocean absorbs more solar irradiance than the ice, creating a positive feedback to warming. Despite the critical role of the Arctic in recent climate change, we have very little information about the links between this ocean/ice world and changing temperatures in the geologic past. It is particularly unclear how changing sea-ice volumes affected the formation and circulation of water masses formed in the Arctic. On page 68 of this issue, Brian Haley and co-authors2 analyse the neodymium isotopic composition of ocean sediments to reconstruct the history of Arctic intermediate water circulation. They show that brines produced during Arctic sea-ice formation controlled the formation and characteristics of intermediate water masses over the last 15 million years.
Much of our understanding of past climate conditions comes from deep-sea sediments that record a relatively continuous record of climate through time and often include biologically derived material that can be used for dating and studies of environmental conditions. The ice-covered Arctic Ocean represents a formidable drilling challenge (see the Backstory on page 76). Nevertheless, in 2006 the Integrated Ocean Drilling Program recovered a >400 m composite of sediment cores from the Lomonosov Ridge spanning 
55 Myr (ref. 3). Unfortunately the material in the upper 200 m (the past 16 Myr) largely comprises mineral, rather than biogenic, sediments, which precludes the use of most proven and widespread analytical techniques such as investigation of oxygen and carbon isotopes of biological carbonates. But iron–manganese coatings that form on sediments accumulating on the sea floor contain measurable quantities of neodynium. Nd isotopes extracted from these coatings are one of the few tracers of past ocean circulation. The isotopic signature of a water mass is imparted from weathering of continental material in the region where the water mass sinks below the surface. Radiogenic Nd isotope values reflect a contribution from relatively young volcanic material, whereas non-radiogenic values are a consequence of inputs from old continental crust. The water mass carries its distinctive signature as it circulates through the oceans.
Deep waters in the Southern Hemisphere form as the brines produced during sea-ice formation increase the density of cold surface water. In contrast, modern North Atlantic deep water, which has distinctly non-radiogenic Nd isotope values, forms as the relatively warm and salty surface waters from the North Atlantic current cool and sink in the Greenland–Iceland–Norwegian Sea. However, the Nd extracted from the Lomonosov Ridge sediments was surprisingly radiogenic. Haley and co-authors suggest that intermediate water formed in the Kara Sea bathed the site as the coatings were deposited (Fig. 1). Owing to input of weathering products from the Putorana Basalts in Siberia, surface water in the Kara Sea region had a radiogenic Nd-isotope value. This signature was subsequently carried to intermediate depths as a brine generated during sea-ice formation. Pacific waters are another potential source of such radiogenic Nd isotopes, but the connection between the Atlantic and Pacific oceans through the Bering Strait was closed until 5.5 Myr ago (ref. 4).
Figure 1: Changing circulation patterns in the Arctic Ocean.
a, Between 12 and 2 million years ago and during glacial periods since then, Arctic Intermediate Water (AIW) was mainly produced in the Kara Sea, bathing the core site (red star) in water with a relatively radiogenic neodynium signature (purple arrows). Ventilation in the North Atlantic Ocean occurred south of the Greenland-Iceland-Norwegian (GIN) seas, forming non-radiogenic Northern Component Water (NCW, blue arrows). Deep-water formation in the GIN seas was greatly reduced. b, During interglacial periods that occurred over the last 2 million years, deep-water formation occurred mainly in the Greenland-Iceland-Norwegian seas, producing non-radiogenic North Atlantic Deep Water (NADW) that flows into both the Atlantic and Arctic Ocean basins. Antarctic Intermediate Water production is reduced, and non-radiogenic North Atlantic Deep Water dominates the core location.
Full size image (22 KB)The scenario Haley and co-authors propose would have required an open Fram Strait to allow inflow of seawater into the Arctic, and the development of sea ice in the Kara Sea at a time for which there is conflicting evidence regarding the existence of substantial ice in the Northern Hemisphere3. Finally, the inflow of North Atlantic intermediate waters over the Greenland–Scotland Ridge must have been limited between 12 and 2 Myr ago, requiring ventilation of North Atlantic deep waters at a location south of the Greenland–Iceland–Norwegian seas. A similar ventilation scenario has been proposed for Pleistocene glacial intervals due to sea-ice coverage in the Greenland–Iceland–Norwegian Sea5, and is supported by more-radiogenic Nd isotope values in the Arctic during cold periods over the past 400,000 years2 (Fig. 1). This correlation confirms the idea that more-radiogenic Nd values are not related to Pacific inflow because the sea level was significantly lower during glacial periods, which severely limited flow from the Pacific into the Arctic Ocean through the Bering Strait.
Although independent analyses of the same Arctic cores support the development of Arctic sea ice as early as 16 Myr ago3, the cause of the required shift of North Atlantic ventilation to a more southerly location between 12 to 2 Myr ago — when the Greenland–Iceland–Norwegian seas were presumably largely ice-free — remains unclear.
The limited data for inferring past conditions in the Arctic Ocean mean that the interpretations are relatively unconstrained. However, Haley and colleagues present an interesting idea that other scientists can evaluate and challenge. An important implication of their work is that brine rejection during sea-ice formation may have been the dominant process in making surface waters dense enough to sink to deep and intermediate water depths during the past 15 Myr in both the Northern and Southern Hemispheres. In that case, modern deep-water formation patterns in the North Atlantic Ocean would represent an anomaly rather than the norm6.
The potentially widespread occurrence of brines in the North Atlantic Ocean has important implications for paleoceanographic studies based on the oxygen isotopes of material derived from these waters. Much of what we have learned about conditions and processes occurring in this region comes from this proxy. Yet interpretations of oxygen isotopes are highly dependent on estimates of seawater salinity7, which may have been very different in a system dominated by brine.
Finally, brine formation in the Arctic Ocean throughout the past 15 Myr implies the presence of persistent sea ice, and therefore suggests that the Arctic may have played a critical role in the cooling of the Northern Hemisphere through the ice-albedo feedback, in a reverse of the climate response that we are observing today.